Natural Language Inference on MSciNLI (ECE and BS)
2.08ECECaliDist (As.)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CaliDist (As.)LLM=QWEN-3 8B2026.06 | 2.08 | 22.76 | |
| CaliDist (Sa.)LLM=QWEN-3 8B2026.06 | 2.88 | 22.55 | |
| CaliDist (Pr.)LLM=LLAMA-3.1 8B2026.06 | 4.98 | 21.82 | |
| CaliDist (As.)LLM=LLAMA-3.1 8B2026.06 | 5.75 | 21.94 | |
| CaliDist (Sa.)LLM=LLAMA-3.1 8B2026.06 | 6.23 | 21.58 | |
| CaliDist (Pr.)LLM=QWEN-3 8B2026.06 | 6.51 | 23.44 | |
| Temperature ScalingLLM=LLAMA-3.1 8B2026.06 | 12.1 | 22.38 | |
| VanillaLLM=LLAMA-3.1 8B2026.06 | 25.76 | 28.3 | |
| EntropyLLM=LLAMA-3.1 8B2026.06 | 28.05 | 30.75 | |
| FSDLLM=LLAMA-3.1 8B2026.06 | 29.09 | 29.09 | |
| ConsistencyLLM=LLAMA-3.1 8B2026.06 | 34.3 | 34.84 | |
| Temperature ScalingLLM=QWEN-3 8B2026.06 | 35.46 | 35.72 | |
| EntropyLLM=QWEN-3 8B2026.06 | 38.23 | 39.07 | |
| VanillaLLM=QWEN-3 8B2026.06 | 38.6 | 38.6 | |
| FSDLLM=QWEN-3 8B2026.06 | 38.87 | 39.38 | |
| ConsistencyLLM=QWEN-3 8B2026.06 | 40.14 | 40.14 |